Your Amazon CPCs Jumped This Spring and You Didn't Touch a Setting. Here's What Changed.

If your ACoS crept up this spring and your Seller Central settings look exactly the same, you’re not imagining it. Amazon changed the auction underneath your campaigns, not the campaigns.

In late April 2026, Amazon reweighted its Sponsored Products dynamic bidding engine to lean harder on placement-level conversion probability. When it predicts a top-of-search slot will convert, it now pushes your bid toward that placement more aggressively than before. Agencies across supplements, home goods, apparel, and electronics accessories are reporting cost-per-click up 18% to 27% versus their Q1 baselines — with conversion flat.

Why your guardrails stopped working

“Down Only” was the cost-control lever everyone leaned on. It only lowers bids, never raises them, so it felt safe. But it lowers bids relative to a baseline the engine sets — and that baseline shifted. Sellers running tightly managed Down Only campaigns reported effective CPCs jumping around 22% in the first two weeks of May. The setting in Seller Central didn’t move. The money going out the door did.

Two forces make it bite. AI search — Rufus, now Alexa for Shopping — is collapsing more shopper intent into fewer top-of-search slots, so more demand is chasing the same head-term real estate. And the engine’s new placement weighting means the automatic lift lands harder exactly where the auction is already hottest.

One honest note on sourcing: the 18–27% and the 22% figures are agency- and seller-reported, not published by Amazon. Measure your own accounts before you treat any category average as yours.

The mechanics Amazon does confirm

If you’re going to rebuild, build on what’s documented, not what’s rumored.

• Dynamic bids (up and down) can raise your bid up to 100% for Top of Search and up to 50% for other placements when Amazon predicts a conversion.
• Placement multipliers run 0% to 900% for both top-of-search and product pages.
• Order of operations: the placement multiplier applies first, then dynamic bidding compounds on top. A $1.00 bid with a 900% top-of-search multiplier and up-and-down bidding can reach roughly $20 for a single click.

That stacking is why an aggressive multiplier plus up-and-down bidding produces eye-watering clicks after the April change. The multiplier and the dynamic lift now push the same direction, harder.

Why you saw it late

Because the cost shows up at the SKU level before it shows up in a blended number. A few hero SKUs quietly burn more per click while conversion holds, and account-wide ACoS averages it out for weeks — until enough SKUs drift that the aggregate finally moves. If you manage to one account ACoS, you’re structurally late to this. You need SKU- and placement-level visibility to catch it in days, not a month.

How to rebuild

Treat late April as a structural reset, not a spike that reverses.

1. Stop using Down Only as your spend ceiling. It doesn’t behave like one anymore.
2. Control cost at the placement level. Test top-of-search multipliers by query and product maturity — top-of-search runs 30% to 70% more than rest-of-search but doesn’t always convert 30% to 70% better. Add the multiplier only where the data earns it.
3. Cut by the query. Move budget off inflated head terms and toward mid-tail where intent is specific and CPCs are saner.
4. Only buy traffic your listing can convert. With clicks more expensive and conversion flat, the cheapest way to protect ACoS is a listing that closes — image, title, A+, reviews, price. Ad efficiency is downstream of conversion now.

The operators protecting margin this spring are the ones who stopped trusting Down Only, pushed cost control down to the placement and query, and tightened conversion so every pricier click has a better shot at paying off.

Rebuilding ad economics around an auction shift like this is the work we run for the brands we operate. If your ACoS moved this spring and you want a diagnosis at the SKU and placement level, a strategy assessment is the fastest way to get one.